Law school enrollment declined drastically between the Great Recession and the start of the COVID-19 pandemic. The number of new first-year law students in 2019 was almost 20% lower than in 2011 (Li et al., 2020). Law schools experienced a brief recovery during the pandemic—the number of new students increased by almost 12% in 2021. Yet, the spike was short-lived, and the legal education bubble popped. In fall 2022, the number of new students dropped by 11%. For the 2023 admissions cycle, law schools were expected to receive similarly low, or potentially fewer, applications than they did in the prior year (Sloan, 2022).

Law school administrative offices are deeply concerned with both enrollment and diversity strategies, perhaps more so than all professional school peers (Espeland & Sauder, 2016). Prior scholarship on legal education suggests that when applications dip, admissions offices admit more Black and Hispanic students as a “survival strategy” to achieve their enrollment targets, which aligns with professed commitments to diversity in legal education (Taylor, 2015, p. 321). When law schools admit more students from minoritized racial groups as part of a budget-driven enrollment management strategy, they seek to balance increasing diversity with limiting class sizes. In other words, they aim to enroll more students from minoritized racial groups, but not so many that they must create new course sections. With this context, one law school professor and former law school dean of admissions asked: “to what extent have law schools already manipulated diversity levels for strategic purposes?” (Taylor, 2015, p. 325).

The makeup of law school cohorts matters for both the experiences of individual students as well as the learning environment for entire classes. Prior scholarship finds that Women of Color are so underrepresented in law schools that their experiences are punctuated by tokenism and feelings of being alienated from their white or male peers (e.g., Lain, 2016; Prince, 2017). The overall quality of legal education is also improved when classroom diversity brings varied perspectives into class discussions (e.g., Dark, 1996). Deo et al., (2010) found that law school instructors who are white men often avoid addressing issues related to diversity, even when those perspectives are relevant to the legal material being covered; instead, they rely on women or racially minoritized students to introduce those points into class conversations. As such, beyond enrollment management, the racial and gender composition of law school cohorts is integral to achieving the educational benefits of diversity in the legal training of all law students, perhaps especially those from majority backgrounds.

The purpose of this paper is to examine the ways that law schools manage enrollments by admitting new cohorts of students that are more—or less—reflective than currently enrolled students. Specifically, we address the following research questions: Do law schools’ current enrollments of racially diverse students influence whom they enroll in incoming cohorts? Does the relationship between racial and gender composition in current enrollments and incoming cohorts vary by law school selectivity (i.e., median LSAT scores)? We analyze data that law schools report to the American Bar Association (ABA) to examine how the percentage of currently enrolled groups of students (e.g., Black women, Black men) predicts the number and the log-transformed percentage of incoming students from different race and gender backgrounds.

Our analytic strategy is framed by the concept of intersectionality (Crenshaw, 1989), which calls for examining the mechanisms through which law schools strategically manage specific combinations of social identities in their enrollments, thereby shaping diversity within these educational institutions. Therefore, we also examine whether LSAT scores moderate the relationship between currently enrolled and newly admitted students of same race and gender subgroups. Our findings suggest that it is important to disaggregate enrollment patterns by these two intersecting identities and that some of the dynamics that perpetuate racial inequities are especially pronounced for Black women at selective law schools.

In the next section, we briefly review literature addressing legal education and law school admissions. Then we discuss our theoretical framework and how it informs our analytical approach. After that, we discuss the data and methods we use to apply the intersectionality framework to analyze law school enrollments. Toward the end of the paper, we present findings and discuss the implications of those findings for improving admissions practice, supporting diversity in law schools and the legal profession, and addressing inequities inherent in selective admissions.

Review of the Literature

The legal profession has historically promoted White cultural norms while excluding racial minorities and women (Bhabha, 2014). Black students could not enroll in most law schools before the 1950s, and the few People of Color who did enroll were not welcome in the profession (Anderson, 2009). Women also faced immense legal and social obstacles to practicing law. Although all states allowed women to practice law by 1920, women did not comprise more than 10% of all law students until the 1970s (Katz et al., 2022). Currently, more than 50% of law students are women, but only 38% of working lawyers are women, and that proportion is even lower in leadership roles at law firms and law schools (ABA, 2022). Despite substantial gains, Black and Hispanic students remain underrepresented in law schools and the legal profession (Nance & Madsen, 2014; Nussbaumer, 2006; Redfield, 2009). In 2019, 31% of incoming law students were from racially minoritized groups (ABA, 2020), which is well below a U.S. population in which 42% identify with a racially minoritized group (i.e., not White non-Hispanic; U.S. Census Bureau, 2020). Black and HispanicFootnote 1 applicants also experience higher shut-out rates (not being admitted to any law school) relative to the total pool of applicants (AccessLex Institute, 2018).

Race, Rankings, and the LSAT

The White-dominant and highly competitive culture of law school may discourage many Students of Color from applying (Anderson, 2009; Clydesdale, 2004), but even when Students of Color do apply, they are often at a disadvantage due to the Law School Admission Test (LSAT). The LSAT is used by almost all ABA-accredited law schools as a central component of the admissions process. It consists of multiple-choice sections on logical reasoning, math games, and reading comprehension and a single essay section, all of which take about four hours to complete. As with other standardized tests used in higher education, the LSAT has been criticized for assessing particular attributes and being structured in ways that advantage White men (Subotnik, 2013). White, Asian, and male test takers have consistently scored higher than other groups over decades of testing (Lauth & Sweeney, 2022; White, 2001). The possible reasons for these score discrepancies include prior education, socioeconomic status, biased questions, and stereotype threat, among other factors (Hill, 2020). Johnson (2013) noted that applicants from minoritized racial groups systematically score lower on the LSAT than White applicants and criticized law schools for over-relying upon or misusing LSAT scores in selective admissions decisions. That said, he argued that the fundamental problem is that LSAT takers from these groups were applying to the wrong law schools (i.e., law schools that are too selective for their lower LSAT scores).

Almost all law schools use a combination of LSAT scores and undergraduate GPAs as the primary factor when determining whether an application is presumptively admitted, denied, or sent for further committee review (Espeland & Sauder, 2016; Randall, 2006). The consequence of this practice is that many applicants’ full files are not evaluated at all. Heavily weighting LSAT scores in admissions is based on the assumption that the test reliably predicts future success. But scholars have pointed out that although the LSAT predicts first-year grades, it is not the best predictor of bar passage or success in the legal profession in general (e.g., Hill, 2020). Better predictors include law school GPA after the first year, experience in a law-related job while in school, and regular participation in class (Taylor et al., 2021). The LSAT fails to measure attributes such as creativity, which can help some students succeed in law school. Fundamentally, the LSAT measures a student’s ability to succeed at test-taking (Gilmore, 2016).

ABA accreditation requires that law schools use some sort of valid and reliable admissions test (ABA, 2023), but this is not the primary reason that law schools over-rely on the LSAT. Law school administrators are incentivized to recruit high-scoring applicants because of their importance in determining national rankings. Rankings establish the prestige of law schools and the type of students that will apply and attend (Espeland & Sauder, 2016). U.S. News and World Report publishes the most influential annual rankings, ostensibly intended to provide transparency during the hyper-competitive application process. Instead, their effect has been to change the behavior of schools themselves, as administrators focus on how to score highly on the somewhat arbitrary ranking system (Espeland & Sauder, 2016; Sauder & Lancaster, 2006). A substantial difference in rankings between schools can be explained by institutional selectivity, which is calculated from the median LSAT score, median undergraduate GPA, and acceptance rate of each school (Morse et al., 2023). Law schools have responded to this particular metric by leaving tenure-track faculty lines unfilled to allocate more money towards recruitment scholarships for applicants with high LSAT scores (Espeland & Sauder, 2016). The rankings have been further criticized for not accounting for racial equity, diversity, and inclusion (Espeland & Sauder, 2009). The end result is that schools receive no reputational benefit for improving racial diversity and may simply strive to maintain a minimally acceptable number of Students of Color (Bowman et al., 2023).

Racialized Financing of Legal Education

The high price of law school also affects the number of Students of Color who enroll. Law students are more likely to take out loans than other graduate students (Pyne & Grodsky, 2020), and they usually borrow larger amounts (Belasco et al., 2014; Pyne & Grodsky, 2020). Between 2011 and 2017, law school sticker prices increased by 17% for public law schools and 15% for private law schools (Whitford, 2018). In 2020, the average amount borrowed for law students who took out loans was $133,480 at private schools and $93,131 at public schools (Law School Transparency, 2021). Black and Hispanic law school graduates end up with more debt than White graduates and have a harder time reducing that debt load in later years (Hanson, 2021). These trends reflect historical and structural inequities that disadvantage Black and Hispanic borrowers across U.S. higher education (Addo & Baker, 2021; Gándara & Zerquera, 2021).

Due to the pressure to recruit students with high LSAT scores and undergraduate GPAs, merit aid is the primary form of financial aid at law schools (Taylor, 2019). Unfortunately, it is difficult to determine who is financially incentivized to enter law school, because law schools do not share detailed information about their admissions processes. Taylor (2019) analyzed results from the 2016 Law School Survey of Student Engagement to show that law schools offered merit scholarships and grants in patterns that systematically disadvantaged Students of Color: 66% of White students received merit aid offers versus only 49% of Black students and 52% of Hispanic students. In a multi-institutional analysis, Bowman et al. (2023) found evidence that law school scholarship offers may vary greatly across racial groups. Increased transparency in disclosures from law schools would clarify the impact of financial aid packages for different groups of students, but it is likely that current financial aid practices cause legal education to be more expensive for People of Color.

Conceptual Framework

The conception of Critical Race Theory (CRT) in the legal field challenged the notion that the legal system upholds race-neutral practices while asserting that the legal system, in fact, conserves White privilege (Valdes et al., 2002). In her delineation of CRT, Crenshaw (1989) posited that “the intersectional experience is greater than the sum of racism and sexism” (p. 140). She further argued that “any analysis that does not take intersectionality into account cannot sufficiently address the particular manner in which Black women are subordinated” (p. 140). Although Crenshaw focused on using the concept of intersectionality to examine the experiences of Black women, other groups of Women of Color embraced the framework to identify and challenge the “complexities of oppression” they endure (Harris & Patton, 2019, p. 350).

Intersectionality is used to “reveal the inextricable connection between racism and patriarchy in the lives of Women of Color as well as in institutions that support hierarchies of power” (Bowman et al., 2006, p. 57). Due to the interlaced linkage between race and gender, the unique experiences of Women of Color are often overlooked when scholars do not consider an intersectionality perspective (Crenshaw, 1989). For example, the legal community celebrated when more women enrolled in U.S. law schools than men, but few took note that Black and Hispanic women were still underrepresented and that law schools were celebrating admitting large percentages of White women applicants (Nance & Madsen, 2014).

Intersectionality and Quantitative Approaches

Scholars have shown that there are statistically significant differences in outcomes for students with at least two intersecting minoritized identities, which are not evident when models examine race and gender statuses separately (e.g., Jang, 2018; López et al., 2018). While it is important to identify differences across subgroups of People of Color, Harris and Patton (2019) caution that approach, alone, represents a misapplication of intersectionality. They suggest that researchers must also consider the “sociohistorical systems and structures of inequality” to achieve “a nuanced analysis that examines the institutional and societal policies, procedures, and programs that (re)create” inequity for Women of Color (Harris & Patton, 2019, p. 363). Harris and Patton (2019) indicate that researchers may use quantitative data and intersectionality to consider whether outcomes vary for Women of Color compared to Men of Color or White students; however, they must consistently consider the sociohistorical context that leads to differences in outcomes, rather than attributing variation in outcomes to the Women of Color themselves.

We draw on the concept of intersectionality to address our research questions and to identify potentially heterogeneous effects of current law school enrollments on incoming enrollments (Stage & Wells, 2014). Guidelines for using statistical methods for intersectionality research discuss using multiplicative interaction terms to examine how a focal variable—in our case, race—is moderated by another identity variable—in our case, gender (Schudde, 2018). Because prior literature provides substantial evidence that use of the LSAT disproportionately excludes People of Color from law school (e.g., Kidder, 2001; Nussbaumer, 2006; Olivas, 2005), we then used interaction terms with median LSAT scores to align with intersectionality theory’s call for considering the sociohistorical context for legal education. Our intersectionality approach follows prior work in legal education, which shows that gender moderated the relationship between race and law school admission; specifically, unlike Black men, Black women “do not receive the full strength of the independent positive relationship between being Black and admission” (Fernandez et al., 2022, p. 474) and that the relative odds of admission vary based on how law schools are ranked and stratified. Through this approach, this paper also seeks to make a methodological contribution by illustrating whether and how an intersectional analysis may shed additional insight into law school enrollment.

Method

Data Source and Analytic Sample

The sample consisted of 187 ABA-accredited law schools in the U.S. that did not open, close, or merge with another law school during the study. The analyses examined ABA data that were released from 2017 to 2020. Prior to 2017, data on the number of entering first-year law students was provided by race and by gender, but not for each race x gender intersectional group, so the analyses could not be conducted for these earlier years. The admissions process in each year within the sample occurred almost entirely before COVID-19 started having substantial effects in the U.S.; the lone exception is that the outcomes in the final year measured new first-year student enrollment in Fall 2020, and the admissions processes at most law schools were generally not finished by March 2020. (Preliminary analyses showed that the substantive findings were unaffected by excluding that final year from the sample.) A total of four years of data were examined for a total of 742 school x year observations; fewer than 1% of all possible observations were omitted as a result of missing data.

Measures

The two types of dependent variables were the percentage and the number of incoming first-year J.D. students from various race x gender groups. Because some law school students do not provide information about their race (Ford et al., 2022), the percentages were calculated based on the number of incoming students for whom racial data was available. Eight race x gender groups were examined as outcomes: Asian men, Asian women, Black men, Black women, Hispanic men, Hispanic women, White men, and White women. The inclusion of both minoritized (e.g., Black women) and privileged groups (e.g., White men) enabled the analyses to consider the consistency and/or variation in relevant dynamics across identities. Here, we note that we use minoritized to refer to the process and result of historical and social classification schemes that marginalize People of Color in society, regardless of numeric representation (see Chase et al., 2014). We did not examine American Indian/Alaska Native students or Native Hawaiian or other Pacific Islander students, since 88–98% of all observations only included zero or one incoming student within each of these race x gender intersectional groups, so there was not sufficient variance to conduct the analyses. Although the sample sizes for multiracial students were relatively larger, it is unclear whether law schools or prospective law students are attentive to the representation of this heterogeneous group. In addition, during the time covered in this study, the ABA collected data that allowed students to report a non-binary gender. However, given that 96% of observations did not include any incoming students who disclosed a non-binary gender, these groups were also not examined as a result of insufficient sample size.

The key independent variables were the percentages of currently enrolled students of each race x gender group. In addition to the eight groups that were modeled as outcomes, two additional variables indicated the percentage of Men of Color and Women of Color from additional racial or national groups (i.e., American Indian/Alaska Native, Native Hawaiian or other Pacific Islander, multiracial, nonresident alien). Similar to the Integrated Postsecondary Education Data System, these ABA data framed international students as a category within race that does not include the actual racial identity of these students. Thus, “nonresident alien” was incorporated within the same group as these additional racially minoritized identities.

A variety of control variables were included based on prior research on the diversification of law school enrollments (Bowman et al., 2022, 2023; Espeland & Sauder, 2016; Kennedy, 2020; Taylor, 2019) as well as the availability of relevant data. Several financial predictors were used to indicate the percentage of students who received grants or scholarships worth at least half of tuition, percentage of students who received grants or scholarships for less than half of tuition, total of listed full-time tuition and fees (using in-state tuition for public institutions), and average cost of off-campus living expenses (as reported by the law school). Additional variables measured the total number of students enrolled, the percentage of law school instructors who hold racially minoritized identities, and the median LSAT score. For our second research question, the median LSAT score variable was used to examine whether enrollment of race x gender subgroups varied as a function of selectivity. For the supplemental gender-only analyses, the numbers of incoming women and men as well as the percentages of incoming women and men served as outcomes, and the percentages of all enrolled women and all enrolled men served as independent variables.

All predictors were lagged by one year relative to the outcome. For instance, when predicting the Fall 2017 enrollment of incoming students for each race x gender group, the control variables indicated law school characteristics during the 2016–2017 academic year so that they reflect attributes during the relevant admissions and recruitment process. Moreover, given the substantial skew of some continuous variables, natural log transformations were conducted for the outcomes for percentages of incoming students for each race x gender group and for predictors that measured the percentage of student enrollment for each race x gender group, percentage of Faculty of Color, and total law school enrollment. Some law schools had 0% representation of some identities and the natural log of zero is undefined, so one was added to each representation variable before computing the natural log. These log transformations ensured that law schools with very high values on the respective variable do not have a disproportionate impact on the findings. Descriptive statistics for all variables appear in the Appendix.

Analyses

As an initial step, detailed descriptive statistics for the count outcomes were produced to provide an overview of the representation of incoming students whose race and gender are minoritized or privileged across U.S. law schools. These statistics include the median, 25th and 75th percentiles, minimum and maximum, skewness, and kurtosis.

Regression analyses that employed school and year fixed effects were conducted. These fixed effects models remove all between-school variance so that the coefficients convey the extent to which within-school changes in the predictors were associated with changes in the outcomes. By accounting for all observed and unobserved differences across law schools, this approach increases the likelihood that the observed results reflect causal estimates of the impact of each predictor on the dependent variable (see Allison, 2009). That said, although fixed effects models have the benefit of avoiding omitted variable bias at the school level, this bias can still occur through the omission of relevant time-variant predictors (Zhang, 2010). Fixed effects analyses are also well-suited for addressing the phenomenon of interest in this study, which considers how law schools and students may respond to within-school changes in student representation over time. These analyses can be summarized via the following equation:

$$y_{it} = \alpha_{it} + \beta {\bf{x}}_{it} + \gamma {\bf{z}}_i + \delta {\bf{w}}_t + \varepsilon_{it}$$

such that \({y}_{it}\) is the outcome variable, \({\mathbf{x}}_{it}\) is a vector of time-varying predictors, \({\mathbf{z}}_{i}\) is a vector of dummy variables representing each law school (leaving out one as the referent group), \({\mathbf{w}}_{t}\) is a vector of dummy variables representing each year (also leaving out one as the referent group), \({\alpha }_{it}\) is the intercept, and \({\varepsilon }_{it}\) is the error term. Across the vast majority of outcomes, 10–40% of the total variance occurred within law schools, thereby indicating a considerable amount of within-school heterogeneity during this study. The two exceptions were that the log-transformed percentages of White men and women were more stable over time, with 5–6% of total variance occurring within schools.

The count outcomes for the number of incoming law students who hold a particular race x gender intersectional identity were modeled using negative binomial regression. This analytic approach appropriately accounts for the fact that these outcomes were overdispersed, such that each variance was greater than the respective mean (Hilbe, 2011). Likelihood ratio tests indicated that these negative binomial regression analyses provided a better fit in every model than Poisson regression analyses, which instead assume that the mean and variance of the dependent variable are identical to each other (ps < 0.001).

We initially explored the use of random effects analyses that simultaneously examine within- and between-school variance; this approach is quite similar to hierarchical linear modeling with grand-mean centered predictors (Cheslock & Rios-Aguilar, 2011). However, Hausman (1978) specification tests indicated that the fixed effects model was preferred to the corresponding random effects model for predicting every outcome (ps < 0.001), so we present only the fixed effects results.

The analyses included virtually all of the predictors discussed above. The lone exception was that one of the percentage race x gender enrollment groups was omitted from the set of predictors as a quasi-reference group (the use of percentages that were then natural log transformed, as opposed to being modeled via several categorical yes/no variables, means that this did not constitute a true reference group). In most instances, the analyses predicting an outcome that involved the representation of incoming women omitted the current percentage of White women as a predictor; similarly, analyses predicting an outcome with incoming men omitted the current percentage of White men. That said, these current enrollments of White students are important within analyses predicting enrollment of White students; in those instances, Black women served as the reference for White women, and Black men served as the reference for White men. These comparisons of Black and White women are consistent with the original conceptions of intersectionality (Crenshaw, 1989). Sixteen analyses predicting the primary relationships of interest were conducted (4 racial groups × 2 gender groups × 2 measurements of the representation of incoming students).

An additional set of analyses explored the extent to which the primary race x gender results were moderated by the median LSAT score of students at the law school. Each of the preceding analyses was conducted with the same predictors, plus two interaction terms: (1) median LSAT x same race and gender combination as the outcome, and (2) median LSAT x same race and different gender as the outcome. For instance, analyses predicting the enrollment of incoming Asian women contained interactions between median LSAT and percentage of currently enrolled Asian women and between median LSAT and the percentage of currently enrolled Asian men. As a robustness check, all models were re-estimated using U.S. News and World Report law school rankings instead of median LSAT score. Median LSAT scores and rankings were very highly correlated within this sample (r = −0.92). Not surprisingly, then, the results were quite similar for models that used median LSAT score versus law school rankings. As an additional robustness check, subgroup analyses divided the sample in half to explore schools with lower versus higher median LSAT scores (rather than conducting an interaction between average LSAT score and students’ racial demographics); this approach avoided the specification of LSAT scores as interacting in a linear manner with the race x gender representation variable. The pattern of these subgroup results was generally consistent with the pattern of interactions reported here.

A final set of analyses sought to consider whether the present findings from the primary analyses were simply attributable to regression to the mean. Specifically, a low or high incoming enrollment of a particular group of students may be the product of random chance, and the next year would likely be closer to the average for that law school. If regression to the mean were playing a large role in the year-to-year changes, then one would expect to find a negative relationship between current enrollment and the next year’s incoming enrollment for any identity group. Therefore, we conducted fixed effects analyses that examined the representation of women and men instead of the eight race x gender groups in the primary study. The race x gender variables were replaced by corresponding gender-only variables, whereas other aspects of the analyses were the same (i.e., separate analyses for the number versus percentage of incoming students, inclusion of the same set of control variables). For instance, the percentage of enrolled women served as the lone key predictor of the number of incoming women in the next year and the percentage of incoming women in the next year. We used gender to conduct these supplemental analyses, because this construct does not receive the considerable attention that race does within the admissions process, and previous research has already identified year-to-year balancing dynamics by race when this was examined as a single variable without exploring its intersection with gender (Bowman et al., 2023).

Limitations

Some limitations should be noted. First, as mentioned earlier, the ABA Standard 509 Information Reports provide codes for nonresident alien and multiracial students as two mutually exclusive options within a list of “racial” categories, so we were not able to determine the specific racial group(s) with which these students identify. If anything, inaccuracies in these classifications are likely to result in error that would diminish the chances that the analyses would identify significant results. Second, although we were able to conduct analyses for eight different combinations of race (Asian, Black, Hispanic, and White) and gender (men and women), the modest representation of American Indian/Alaska Native students, Native Hawaiian or other Pacific Islander students, and gender non-binary in U.S. law schools prevented us from doing so for groups that included these identities. Third, the analyses were limited to school-level information that we were able to obtain through public data. For instance, we ideally would have created variables that indicated the representation of race x gender identities for faculty at each law school, but the lack of available information led us to create a variable indicating the percentage of all Faculty of Color. Fourth, these data provide insights into the behavior of law school admissions offices, but we do not have direct evidence about their decision making.

Finally, the analyses contained a modest number of observations per law school as a result of data limitations before 2017 and the proliferation of COVID-19 in recent years, which could substantially affect the dynamics underlying changes in student representation. Despite having less than ideal statistical power, numerous significant results pertaining to the relationships of interest were identified. Given these modest sample sizes, results are displayed for different levels of statistical significance (including p < 0.10), so readers can choose the level that they consider appropriate. To further contextualize the findings, the summary and discussion of results below highlights differences in the strength of relationships between variables and the significance value of specific results. Fortunately, the primary patterns of findings are unchanged regardless of the criterion used to determine statistical significance.

Results

Table 1 contains detailed descriptive statistics for the number of incoming law students. These entering cohorts tend to be small for each racially minoritized group by gender, with a median of nine Hispanic women, seven Black women, seven Hispanic men, four Black men, four Asian women, and three Asian men. These values contrast considerably with a median of 57 White women and 56 White men. The distributions for all groups of racially minoritized students are considerably right skewed, with skewness values ranging from 2.08 to 4.68 and maximum values that are 5–15 times larger than the corresponding 75th percentile. Figure 1 provides an illustrative example of a histogram for the number of incoming Black women.

Table 1 Detailed descriptive statistics for the number of incoming law students by race and gender
Full size table
Fig. 1
figure 1

Histogram of the number of incoming Black women students at all law schools

Full size image

The findings for fixed effects regression analyses predicting the number of incoming students in each race x gender group appear in Table 2. In all eight analyses, an increase in the representation of a particular race x gender group in one year predicts a decrease in that same group within the following year. These results are highly significant (ps < 0.001) in seven of the eight analyses, with White women being the lone exception (p < 0.05). In contrast, an increase in the representation of a particular race x gender group generally leads to a subsequent increase in incoming students who had the same race but a different gender. For instance, the percentage of currently enrolled Black men predicts a greater number of incoming Black women; the same is true for the percentage of currently enrolled Black women predicting greater enrollment of incoming Black men. The two nonsignificant results are for currently enrolled Hispanic men predicting incoming Hispanic women and currently enrolled Asian men predicting incoming Asian women. That said, even when these different-gender results are significant, the magnitude of the relationship is often much smaller than for the same-gender patterns in the opposite direction, with the notable exception of White students (for whom these two pairs of relationships are nearly identical in size).

Table 2 Unstandardized coefficients for negative binomial fixed effects regression analyses predicting the number of incoming law school students by race and gender
Full size table

When examining predictors involving representation of students from different racial groups, few results are statistically significant, especially at p < 0.05. The percentage of currently enrolled Black women is positively associated with the incoming enrollment of Hispanic women and Hispanic men. In two instances of same gender and different race, the current enrollment of Asian women is positively related to enrollment of Hispanic women, and current enrollment of Asian men is positively related to enrollment of Hispanic men. Moreover, the current enrollment of White men predicts fewer incoming Asian women, whereas the current enrollment of Asian women predicts greater enrollment of incoming White men. In a pair of consistent results at p < 0.10, the percentage of Men of Color from other races predicts decreased enrollment of Black women and Black men in the next year.

As shown in Table 3, similar patterns are observed when predicting the log-transformed percentage of incoming students for each race x gender group. In all eight analyses, increases in the current enrollment of each group predict decreases in the percentage of that same group within the next year (ps < 0.001). Increases in one race x gender group are sometimes associated with a greater representation of incoming students from the same race and different gender, but these findings are less consistently significant for this set of percentage outcomes than for the count outcomes. Specifically, the enrollment of White men predicts increased subsequent enrollment of White women, current enrollment of White women predicts increased subsequent enrollment of White men, and enrollment of Hispanic women predicts greater subsequent enrollment of Hispanic men at p < 0.05; the current enrollment of Black women also predicts greater incoming representation of Black men at p < 0.10.

Table 3 Unstandardized coefficients for fixed effects regression analyses predicting the log-transformed percentage of incoming law school students by race and gender
Full size table

Similar to the findings for the count outcomes, few results are significant for different racial groups. In a notable exception, the current enrollment of Black women predicts increased subsequent enrollment of both Hispanic women and Hispanic men. The current representation of Asian men predicts greater enrollment of Hispanic men, and the current enrollment of Hispanic men also predicts greater enrollment of Asian women. No other results are significant at p < 0.05, and only a few additional scattered results are significant at p < 0.10.

Table 4 displays the results for fixed effects analyses that contain interactions between key current representation variables and median LSAT score. For both types of outcomes (total number and percentage of students), the inverse relationship between the current enrollment of Black women and the prevalence of incoming Black women is stronger (i.e., more negative) at law schools with higher LSAT scores. In addition, the positive relationship between the current enrollment of Black women and the percentage of incoming Black men is more modest (i.e., less positive) at law schools with higher average LSAT scores. In the opposite direction of the findings for Black women, the inverse association between the current enrollment of White men and the incoming percentage of White men is more modest (i.e., less negative) at law schools with higher LSAT scores. These four moderation results are all significant at p < 0.05; no other interaction is significant even at p < 0.10.

Table 4 Unstandardized coefficients for fixed effects regression analyses of the interaction between the median LSAT score and current student representation predicting incoming law school student enrollment by race and gender
Full size table

Finally, supplementary fixed effects regression analyses examined the extent to which current enrollments of women or men predict the number and percentage of incoming students in the following year who hold the same gender as an approach for considering whether the intersectional findings are simply a form of regression to the mean. Contrary to the highly consistent findings by race and gender, the current enrollment of women was not significantly related to the number of incoming women in the following year (B = −0.103, SE = 0.201, p = 0.61), and the corresponding relationship for the current enrollment of men predicting the number of incoming men in the next year was also nonsignificant (B = −0.102, SE = 0.212, p = 0.63). When examining the percentage-based outcomes, these relationships were significant for predicting the subsequent percentage of incoming women (B = −0.181, SE = 0.049, p < 0.001) and percentage of incoming men (B = −0.198, SE = 0.052, p < 0.001).

Discussion and Implications

Prior scholarship indicates that law schools are strategically diverse. They seek to admit racially minoritized students, especially when they are worried about meeting enrollment targets, without overenrolling students and needing to open more course sections (e.g., Taylor, 2015). People of Color are underrepresented in legal education, and research has largely attributed this problem to use of the LSAT in admissions decisions (e.g., Hill, 2020; Kidder, 2001; Nussbaumer, 2006; Olivas, 2005). We sought to build on prior work by examining whether law schools may actually be practicing an enrollment management scheme where diversity of incoming cohorts is inversely related to the diversity of currently enrolled students.

Informed by the concept of intersectionality, we offer new empirical findings about how law schools manage enrollment of People of Color by balancing gender diversity. In other words, we found that it is important to examine the intersection of race and gender to understand the ways in which law schools manage diversity. Across subgroups (e.g., Black women, White men), law schools offset current enrollments with fewer students of that same intersectional group in future cohorts. Additionally, rather than admit more Black men and Black women, our findings indicate that law schools limit overall Black enrollment; when the current enrollment of Black women is higher, incoming cohorts include fewer Black women and more Black men. Similarly, law schools balanced Hispanic women law students’ enrollment with Hispanic men. These results were robust across fixed effects models predicting both the number and the percentage of incoming students. Importantly, these findings for differential results by race and gender could not have been obtained through analyses that focus on a single identity group. Single-group analyses would overlook the fact that the dynamic of racial balancing over time actually occurs for each race x gender group, whereas the reverse pattern is sometimes observed for same race and different gender groups.

We also examined whether selectivity in law school admissions, particularly as measured by median LSAT scores, is associated with this approach to enrollment management. Scholars have previously criticized the use of the LSAT in selective admissions decisions (e.g., Kidder, 2001) and how use of the LSAT matters for law school rankings and prestige (e.g., Sauder & Espeland, 2007). Building on prior discussions on use of the LSAT, we showed that the pattern of findings—especially related to Black women—is most prevalent at selective law schools. That is, the negative relationship between current enrollment of Black women and incoming enrollment of Black women was especially pronounced at law schools with higher median LSAT scores. Our initial results showed that current enrollment of Black women was generally related to increasing future enrollment of Black men; however, this positive relationship for Black men was not as large at law schools with higher median LSAT scores. Finally, we found some evidence (albeit less consistent) that the potential negative impact of current White male enrollments on the incoming enrollment of White men was smallest at law schools with high median LSAT scores.

Taken together, our findings reveal that the unique combinations of race and gender matter in terms of opportunity to pursue legal education, specifically through processes which continue to render People of Color—especially Women of Color—as perpetually underrepresented in the law school context. Previous work has highlighted how the persistent underrepresentation of Women of Color in law schools leads to tokenism and alienation (e.g., Lain, 2016; Prince, 2017) and limits the places the burden for discussing diversity in the classroom on the few Women of Color who are present (e.g., Deo et al., 2010). Using more recent data, we show how enrollment management practices have perpetuated that underrepresentation for Women of Color in ways that were not evident when only examining gender without race.

Based on our review of the literature and intersectionality framework, we show that the use of the LSAT moderates the relationship between current and incoming enrollments in ways that disadvantage Black students. Thus, the overreliance on the LSAT in selective admissions—and to stratify the market for legal education—seems to disadvantage Black women’s “individual experiences…within mutually constitutive sociohistorical systems and structures of inequality” (Harris & Patton, 2019, p. 363). As mentioned above, robustness checks showed that these conclusions were the same when using law school rankings instead of median LSAT scores. Taken together, the primary models and robustness checks suggest that it is the stratification of law schools that disadvantages Black women, in particular, and that the LSAT is the proxy that legitimizes stratification and exclusivity (see, e.g., Sauder & Espeland, 2007).

Our results partially support—and partially contradict—Johnson’s (2013) conclusions about the use of the LSAT and the lack of diversity in legal education. Our findings suggest that research and practice focusing on applicants, and where they apply to law school, may not be as important to diversifying legal education as challenging the ways that law schools balance out gender and racial diversity. This structural lens encourages more research and institutional reform geared toward strategic enrollment management practices, with a particular eye on intersectional inequities that may be perpetuated through law school administrative offices. We also build on Fernandez and colleagues’ (2022) findings on law school admissions by showing that the inequities related to representational balancing in legal education are worse for Black women at more selective law schools (i.e., those with higher median LSAT scores).

Our findings offer important implications for the use of intersectionality as a framework for research in higher education. Using a panel of national data, we demonstrate empirically that gender can have important moderating effects on race in law school enrollment and that these intersectional patterns are not simply attributable to regression to the mean (as indicated by the lack of consistent findings for gender alone). We then show that law schools with higher LSAT scores—something that has long stratified law school admissions and contributed to the hierarchy of law schools—are particularly likely to strategically manage diversity of Black students while reproducing inequity in favor of White men. Following the U.S. Supreme Court’s 2023 rulings, which put increased scrutiny on how admissions professionals can consider applicants’ race (Orfield & Gándara, 2023), these quantitative findings provide new evidence that schools must consider gender equity, in addition to racial diversity, and also how gender intersects with race. As critics of race-conscious admissions call for further reliance on ostensibly race-neutral standardized tests, we add to prior literature that shows how testing regimes can uphold “structures of inequality” (Harris & Patton, 2019, p. 363).

Future research may apply similar approaches (intersectionality, quantitative analyses) to examine admissions and enrollment decisions, along with how financial aid is awarded, in other professional, disciplinary, and institutional contexts. Additionally, scholars should consider that intersectionality does not need to be limited to gender and race, as other studies can apply the framework to include social identity aspects that contribute to various layers of oppression. Intersectionality and other critical theories can be adopted to consider multiple intersections based on other statuses or identities (e.g., race, gender, class, sexuality, ability status, etc.), forms of systemic oppressions (e.g., racism, sexism, classism, homophobia, ableism, etc.), and the intricate interdependence of those intersections and oppressions at the micro and macro level (Dhamoon & Hankivsky, 2011). Due to limitations in ABA data, we were not able to examine additional gender identities. However, future research may use newer ABA data to further explore enrollment of students who declare a gender identity other than man or woman (Ford et al., 2021). Moreover, although it fell outside the scope of our research questions, other researchers may consider ways to analyze outcomes for students whose race is not known (e.g., Ford et al., 2020).

Finally, future research should seek to provide more direct insight into the decision making of law school admissions offices. A straightforward approach could involve conducting interviews with law school administrators, which would require admissions professionals to be candid about their behavior. Another approach could involve a simulation study that provides admissions officers with statistics for law school enrollments that are disaggregated by race and gender; these admissions officers could be randomly assigned different imbalances by race and gender and then asked who they would seek to recruit in the next upcoming cohort. They could also be asked whether and how the enrollment statistics that they review in their regular practice may diverge from those provided within the simulation study.

Conclusion

By examining national enrollment data for a four-year period, we argue that law schools may be strategically diverse in how they balance overall racial diversity across cohorts based on gender. As a result, the system of selective admissions and law school rankings may directly facilitate a lack of racial and gender equity. Some elite law schools have started withdrawing from participation in—or challenging metrics used for—national rankings (Hartocollis & Fawcett, 2022). However, unless they revisit their own enrollment management approaches, the most selective law schools will continue to limit progress toward diversifying legal education. The present study shed light into these dynamics; it also illustrated how an intersectional analysis can uncover enrollment patterns that would otherwise be overlooked.